Incidence and Determinants of 1-Month Mortality after Cancer-Directed Surgery

Transcription

1 Annals of Oncology Advance Access published November 27, Incidence and Determinants of 1-Month Mortality after Cancer-Directed Surgery B.A. Mahal 1,*, G. Inverso 2,*, A.A. Aizer 3, D.R. Ziehr 1, A.S. Hyatt 3, T.K. Choueiri 4, K.E. Hoffman 5, J.C. Hu 6, C.J. Beard 3, A.V. D Amico 3, N.E. Martin 3, P.F. Orio III 3, Q-D. Trinh 7, P.L. Nguyen 3 1 Harvard Medical School, Boston MA, USA 2 Harvard School of Dental Medicine, Boston MA, USA 3 Department of Radiation Oncology 4 Department of Medical Oncology, Dana-Farber Cancer Institute and Brigham and Women s Hospital, Harvard Medical School, Boston MA, USA 5 Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston TX, USA 6 Department of Urology, UCLA Medical Center, Los Angeles CA, USA 7 Division of Urology, Brigham and Women s Hospital, Harvard Medical School, Boston MA, USA Corresponding Author: Dr. Paul L. Nguyen, Dana Farber Cancer Institute/Brigham and Women s Hospital, Harvard Medical School, 75 Francis Street, Boston, MA, 02115, USA, Phone: / Fax: / pnguyen@lroc.harvard.edu * These authors contributed equally The Author Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please journals.permissions@oup.com.

2 2 Abstract Background: Death within 1 month of surgery is considered treatment related and serves as an important healthcare quality metric. We sought to identify the incidence of and factors associated with 1-month mortality after cancer-directed surgery. Patients and Methods: We used the Surveillance, Epidemiology and End Results Program to study a cohort of 1,110,236 patients diagnosed from with cancers that are among the 10 most common or most fatal who received cancer-directed surgery. Multivariable logistic regression analyses were used to identify factors associated with 1-month mortality after cancerdirected surgery. Results: 53,498 patients (4.8%) died within 1 month of cancer-directed surgery. Patients who were married, insured, or who had a top 50th percentile income or educational status had lower odds of 1-month mortality from cancer-directed surgery ([adjusted odds ratio (AOR) 0.80; 95% CI ; P], [AOR 0.88; ( ); P], [AOR 0.95; ( ); P], and [AOR 0.98; ( ); P=0.043], respectively). Patients who were non-white minority, male, or older (per year increase), or who had advanced tumor stage 4 disease all had a higher risk of 1-month mortality after cancer-directed surgery, with AORs of 1.13 ( ), P; 1.11 ( ), P; 1.02 ( ), P; and 1.89 ( ), P respectively. Conclusions: Unmarried, uninsured, non-white, male, older, less educated, and poorer patients were all at a significantly higher risk for death within 1 month of cancer-directed surgery. Efforts to reduce 1-month surgical mortality and eliminate sociodemographic disparities in this adverse outcome could significantly improve survival among patients with cancer.

3 3 Key Words: Surgery; Mortality Rate; Outcomes Research; Health Policy; Socioeconomic Disparities; SEER Program Abbreviations: American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP), Surveillance Epidemiology and End Results Program (SEER), Key Message: Nearly 1 in 20 patients receiving cancer-directed surgery for the most common or deadly cancers die within 1 month of the procedure, with unmarried, uninsured, non-white, male, older, less educated, and poorer patients at a significantly higher risk for 1-month mortality. Efforts to reduce 1-month surgical mortality and eliminate sociodemographic disparities in this adverse outcome have the potential to significantly improve survival among patients with cancer.

4 4 Introduction: In 2014, it is estimated that there will be 1,665,540 new diagnoses of cancer and 585,720 deaths due to cancer in the United States, alone.[1] Surgery remains a mainstay definitive treatment for cancer and is the gold standard for treatment across many major cancer sites.[2] One-month mortality is used an important surgical quality metric and deaths within 1 month of the date of surgery are considered treatment related when quantifying operative mortality.[3] Reducing 1-month surgical mortality is a priority for providers and policymakers, and programs such as the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) have been developed to help achieve this goal.[4] Nevertheless, there is little data in the literature which elicits the incidence and determinants of 1-month mortality after cancer directed surgery across all major cancer sites. Furthermore, although there is literature to suggest that there are disparities in surgical outcomes negatively impacting disadvantaged populations, few studies have examined the associations between sociodemographics and surgical outcomes among patients with cancer.[5, 6] Identifying the nationwide incidence of and factors associated with 1-month mortality after cancer-directed surgery could inform policy decisions and allow for targeted interventions that aim to improve cancer survival and reduce sociodemographic disparities in adverse outcomes. We used the national Surveillance, Epidemiology, and End Results (SEER) database to study a contemporary cohort of over 1.1 million patients to establish and determine the incidence and determinants of 1-month mortality after cancer-directed surgery for the 10 most common or most fatal cancers in the United States.

5 5 Patients and Methods: Patient Population and Study Design Sponsored by the National Cancer Institute, the Surveillance, Epidemiology and End Results Program (SEER) program collects and publishes cancer incidence, survival, and treatment data from population based cancer registries; the 17 tumor registries encompass nearly 28% of the US population and capture approximately 97% of incident cancers.[7] Using the most contemporary release of the SEER database, we selected and studied a cohort of 1,110,236 patients diagnosed from with localized non-metastatic forms of cancers that are among the top 10 most commonly diagnosed or the 10 which cause the most cancer deaths annually in the US (resulting in 15 total cancer sites: colorectal, esophagus, thyroid, breast, pancreatic, endometrial, ovarian, head/neck, prostate, lung, liver/intrahepatic bile duct, bladder, melanoma, non-hodgkin lymphoma, and kidney cancer)[1] who received cancer-directed surgery of the primary site as a part of curative or radical treatment. Patients were not selected for the study if a diagnosis of cancer was made at autopsy, a prior malignancy was diagnosed, metastatic disease was present, one of the independent covariates defined below was missing, or if patients were not identified as having a surgical procedure categorized by SEER as cancerdirected. The primary outcome of interest was death within 1 month of cancer-directed surgery. We were unable to use 30-day mortality following surgery since SEER codes survival time using month and year (based on dates of diagnosis and death) rather than reporting exact dates. However, given that SEER coding rules dictate that the date of diagnosis must precede any cancer-directed therapy (including surgery), we considered patients who died within the same

6 6 month as diagnosis or in the subsequent month to have died within 1 month of surgery in our primary analyses. This method for obtaining 1-month mortality after cancer-directed surgery from the SEER database has previously been described.[8] The independent covariates of interest were sex, race, marital status, income, educational status, residence, insurance status and tumor stage. Tumor stage was determined using the AJCC 6 th edition as provided by SEER.[7] Sex was classified as male vs female, race was classified as non-hispanic white vs non-white minority (African American, Hispanic/Latino, Asian/Pacific Islander, Native American, or other race), and marital status was classified as married vs not married as designated by the SEER program. Income (computed as median household income) and educational status (computed as the percentage of residents >25 years of age with at least a high school education) were approximated through use of county level data by linking to the 2000 United States Census.[9] Both income and educational status were stratified by respective medians, creating high income (top 50 th percentile) vs low income (bottom 50 th percentile) and high educational status (top 50 th percentile) vs low educational status (bottom50 th percentile) variables. Residence type was also determined by linking to the 2003 United States Department of Agriculture rural-urban continuum codes.[10] We analyzed insurance coverage as a dichotomous variable given that SEER does not provide information on the specific type of insurance coverage that patients have. Specifically, a patient was considered insured if s/he was classified by SEER as insured, insured/no specifics, or any Medicaid, and patients were considered uninsured if they were classified as such. Of note, the insurance variable was only available from as provided by SEER and so the primary analyses capturing information back to 2004 were not able to include insurance status. Furthermore, all analyses detailed below that included insurance status had study populations limited to age 65 as SEER

7 7 recommends exercising caution when using the insurance status variable among patients over 65 given that many who are classified as uninsured are Medicare eligible. Statistical Analysis Baseline patient characteristics for patients who suffered from 1-month mortality after cancer-directed surgery vs those who survived beyond 1 month were compared with the t test or χ 2 test as appropriate. Multivariable logistic regression analyses were used to determine whether there were any determinants associated with 1-month mortality after cancer-directed surgery for all cases diagnosed Sex, race, marital status, income, educational status, residence, and tumor stage were included in the primary multivariable logistic regression analysis. This analysis was performed for all cancers and then repeated by each site. After adjusting for all of the previously listed covariates, multivariable logistic regression analysis was also used to determine whether insurance status was associated with 1-month mortality after cancer-directed surgery for cases diagnosed from 2007 (introduction of insurance variable) 2011 among patients younger than 65, as detailed above. Again, this analysis was performed for all cancers and then repeated by each site. To minimize the potential error in our approach of designating patients who died within one month of surgery (i.e. a patient that was diagnosed and treated on the first day of one month who died on the last day of the next month), we completed sensitivity analyses where we very strictly redefined death within one month to be death within the same month as diagnosis (and not the subsequent month); these patients were diagnosed with cancer, received cancer-directed surgery, and died all in the same month. Multivariable logistic regression sensitivity analyses

8 8 using this strict definition was repeated for all cancers. Furthermore, since cancer-directed surgeries for non-hodgkin lymphoma are rare given that systemic therapy is the mainstay for the disease, we completed a multivariable logistic regression sensitivity analysis excluding all patients diagnosed with non-hodgkin lymphoma (N = 30,669). All P values were two sided. The threshold of 0.05 was used to determine statistical significance. Statistical analyses were performed using STATA 13.0 (StataCorp, College Station, TX) for all analyses. This study was approved by the Dana-Farber/Harvard Cancer Center institutional review board; approval was exempt. Results: Patient Characteristics Baseline clinical and demographic characteristics are displayed in Table 1. Overall, 53,498 patients (4.8%) died within 1 month of cancer-directed surgery; one-month raw unadjusted mortality rates by site are displayed in supplementary table 1. Notably, patients who suffered from 1-month mortality after cancer-directed surgery were 5 years older (on average) than patients who survived at least 1 month after cancer-surgery and were more likely to come from counties with lower income and education levels (P in all cases), although differences were small in clinical magnitude. Furthermore, there was a larger proportion of patients with advanced T3 or T4 disease, males, non-white minorities, and rural residents among those who suffered 1-month mortality after cancer-directed surgery than among patients who survived beyond 1 month (P in all cases).

9 9 Factors Associated With 1-Month Mortality after Cancer-Directed Surgery After adjustment for all sociodemographics, patients who were married, insured, or who had a top 50th percentile income or educational status all had lower odds of 1-month mortality from cancer-directed surgery ([adjusted odds ratio (AOR) 0.80; 95% CI ; P], [AOR 0.88; ( ); P], [AOR 0.95; ( ); P], and [AOR 0.98; ( ); P=0.043]) compared to patients who were unmarried, uninsured, or who had the bottom 50th percentile income or educational status, respectively (Table 2). Patients who were non-white minority, male, or older (per year increase in age) or who had Tumor stage 4 disease all had a higher risk of 1-month mortality after cancer-directed surgery, with AORs of 1.13 (95% CI ; P), 1.11 (95% CI ; P), 1.02 (95% CI ; P), and 1.89 (95% CI ; P), compared to non-hispanic whites, females, younger patients, and patients with T1 disease, respectively (Table 2). Furthermore, when breaking down race, compared to non-hispanic whites, African Americans (AOR 1.17 [95% CI ]; P ), Hispanics (AOR 1.15 [95% CI ]; P ), Native Americans (AOR 1.21 [95% CI ]; P=0.002), Asians (AOR 1.04 [95% CI ]; P=0.025), and other races (AOR 1.11 [95% CI ]; P=0.008) all had an increased risk for death within 1 month of operative procedure. On sensitivity analysis, when 1-month mortality after cancer-directed surgery was very strictly defined as death within the same month as diagnosis and surgery (and not in the subsequent month), the rate of 1-month mortality dropped to 3.3% (N=36,273; Table 3) and all associations remained nearly identical except for the protective associations between both insurance status and the top 50 th percentile of educational status and 1-month mortality after

10 10 cancer-surgery (AOR 0.92 [95% CI ; P=0.05] and AOR 0.99 [95% CI ; P=0.35], respectively) [Table 2]. Furthermore, when excluding patients with non-hodgkin lymphoma (N=30,669), all of the aforementioned factors remained significantly associated with 1-month mortality after cancer-surgery except for top 50 th percentile educational status (AOR 0.98 [ 95% CI ; P=0.21]) [Table 2]. Figure 1 displays forest plots for each of the sociodemographic determinants which were found to be significantly associated with death after 1 month mortality of surgery (marital status, insurance status, race, income, and educational status) by cancer site. These determinants remained significantly associated with 1-month mortality after cancer-directed surgery for most but not all sites (Figure 1 [a-f]). Discussion: Using a large national and contemporary cohort of over 1.1 million patients who underwent cancer-directed surgery, we found that 4.8% of patients died within 1 month of the operation. Patients who were married, insured, or who had a top 50 th percentile income or educational status had 20%, 12%, 5%, and 2% lower odds of suffering from 1-month mortality from cancer-directed surgery, compared to patients who were not married, uninsured, or who had a bottom 50 th percentile income or educational status, respectively. Furthermore, patients who were non-white minority, male, or older (per year increase), or who had T4 disease had 13%, 11%, 2%, and 89% higher odds of suffering from 1 month of cancer-directed surgery compared to whites, females, and younger patients, and those with T1 disease, respectively. These associations remained nearly identical when strictly defining 1-month mortality from cancer-

11 11 directed surgery as death within the same month as diagnosis (except for insurance status [P = 0.05] and educational status [P=0.35]). Furthermore, these associations remained significant within most of the individual cancer sites evaluated. This study is the largest of its kind and draws attention to the concerning finding that nationwide, nearly 1 in 20 patients receiving cancer-directed surgery died within 1 month of the procedure. This result is higher than what has typically been reported in the literature, such as % after radical prostatectomy,[11] approximately 3% after cystectomy,[12] and up to 8.1% after colorectal surgery,[13] but it is important to remember that previously published results mostly come from high-volume specialized academic centers, while the data reported here represent what is going on nationwide. There is mature literature which has demonstrated that 1-month mortality after cancer-directed surgery is closely linked to hospital and surgeon volume; notably, it has been shown that low-volume hospitals have up to a 77% increased relative risk of 1-month mortality compared to high-volume centers after radical prostatectomy and absolute 1-month mortality rates can vary by nearly 30% (from 3.5% to 44.1%) after surgery for colorectal cancer.[11, 14] Furthermore, systems characteristics (i.e. high nurse to patient ratios, presence of complex medical oncology services, presence of positron emission tomography scanners) and hospital complexity (based on degree of clinical variety managed by hospital) have also been demonstrated to be independently associated with surgical mortality rates, with better outcomes in surgical centers with more resources and increasing complexity.[15] High volume and quality surgical centers are able to achieve low 1-month surgical mortality rates likely due to their ability to better protect patients from postoperative complications, as is demonstrated by lower readmission rates among high performing centers.[16] Given our results, it is clear that efforts need to be undertaken to reduce 1-month

12 12 surgical mortality after cancer surgery and interventions should start by identifying underperforming hospitals which may be low-volume, low-complexity centers with limited resources. As has been previously postulated, increasing resources and expanding systems capabilities at low-volume centers could lead to improved surgical outcomes. Our study also uniquely highlights the significant impact that sociodemographic determinants may have on surgical mortality outcomes among patients with the most common or deadly malignancies. The reasons for the observed sociodemographic disparities in 1-month mortality are likely layered and multifactorial. The results of this study may reflect differences in access to high quality and high volume centers, biology/genetics, social support, and differential treatment in hospitals.[5, 17] Furthermore, some authors have suggested that socidemographic characteristics alone account for differences in surgical outcomes,[18] while others have suggested that differences in hospital quality and systems level factors contribute to disparities in outcomes.[17, 19, 20] Although there is little consensus as to why disparities in surgical outcomes exist, most of the existing literature is congruent with respect to one particularly alarming finding: minority, low SES, uninsured, disadvantaged patients disproportionately receive care at lower volume/lower quality, underperforming centers.[17, 18] We hypothesize that disproportionate access to quality health care is the major driver of the disparities in 1-month cancer-directed surgical mortality observed in this study. The findings that unmarried, uninsured, non-white, male, older, less educated, and poorer patients were all at a significantly higher risk for death within 1 month of cancer-directed surgery raise the possibility that poor outcomes after surgery may be a contributor to the poorer cancerspecific and overall mortality observed among these disadvantaged groups.[1] Although these results are congruent with other studies that have evaluated the associations between

13 13 sociodemographics and surgical outcomes after cardiovascular and oncologic operations among Medicare patients,[18, 21] our study is unique in that it also included a large number of patients (N=659,473; 59.4% of the study cohort) younger than 65 who are not yet Medicare eligible and demonstrates that these disparities are still highly prevalent among younger populations. Future interventions aimed at reducing surgical mortality and improving surgical outcomes among cancer patients should target these particularly vulnerable groups and the drivers of these disparities. Our results must be viewed within the limitations of the study. First, given the SEER coding methods we were not able to precisely define 1-month surgical mortality and so our method of considering patients who died within the same month as diagnosis or in the subsequent month to have died within 1 month of surgery in our primary analyses may have included some patients that died up to 2 months after surgery. However, this extreme case scenario would require a patient to be diagnosed and treated on the first day of one month and die on the last day of the next month; statistically, most cases would be diagnosed around the middle of the month with some delay to surgery.[8] Nevertheless, after restricting analyses to death within the same month of diagnosis and surgery, the rate of 1-month mortality only decreased from 4.8% to 3.3% and all associations remained nearly identical except for the protective association between the top 50 th percentile of educational status and 1-month mortality after cancer-surgery (Table 2). Second, SEER does not provide information on comorbidity status or peri-operative risk and so some the difference in 1-month mortality after cancerdirected surgery might have been due to differences in comorbidity or peri-operative risk. However, we were able to control for age which is a significant predictor of surgical mortality and treatment selection should have taken comorbidity into account thereby reducing the

14 14 likelihood that patients with adverse comorbidities were included in this cohort.[22] Third, there may have been delays between diagnosis and receipt of surgery among some patients (especially among those who received neoadjuvant therapies) preventing us from capturing all mortalities occurring within 1 month of surgery, since SEER only provides time from diagnosis to death or last follow-up. Yet, the 1-month mortality rates which we detected were near or above (and on the same order of) those reported from top academic centers. Lastly, SEER does not provide site of care information and we were therefore unable to analyze the data by volume or quality of center. Of note, linkage of SEER to Medicare may be able to address some of the limitations above, however the results would have limited generalizability since SEER-Medicare studies only include patients who are eligible for Medicare. Despite these potential limitations, our study indicates that cancer patients from disadvantaged groups, namely unmarried, uninsured, non-white, male, older, less educated, and poorer patients, were all at a significantly higher risk for death within 1 month of cancer-directed surgery. Policy makers and health care providers should be aware of the poorer outcomes after cancer-directed surgery experienced among these disadvantaged groups and future interventions should target improving access to high quality/volume care. Efforts to reduce 1-month surgical mortality and eliminate sociodemographic disparities in this adverse outcome have the potential to significantly improve survival among patients with cancer.

15 15 Acknowledgements: We are grateful to Dr. Ashish Jha (Harvard School of Public Health) for his comments on this manuscript. Funding: This work was supported by David and Cynthia Chapin, the Prostate Cancer Foundation, Fitz s Cancer Warriors, Hugh Simons in honor of Frank and Anne Simons, and a grant from an anonymous family foundation. There are no grant numbers to be reported. Disclosure: The authors have declared no conflicts of interest.

16 16 References: 1. Siegel R, Ma J, Zou Z, Jemal A. Cancer statistics, CA Cancer J Clin 2014; 64: NCCN Clinical Practice Guidelines in Oncology. Available at: Accessed 30 December Semel ME, Lipsitz SR, Funk LM et al. Rates and patterns of death after surgery in the United States, 1996 and Surgery 2012; 151: Ko C. ACS NSQIP Conference and Semiannual Report Overview. Presentation at the 2009 ACS NSQIP National Conference Lucas FL, Stukel TA, Morris AM et al. Race and surgical mortality in the United States. Ann Surg 2006; 243: Reames BN, Birkmeyer NJ, Dimick JB, Ghaferi AA. Socioeconomic Disparities in Mortality After Cancer Surgery: Failure to Rescue. JAMA Surg Surveillance, Epidemiology, and End Results (SEER) Program: Research Data ( ), National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch, based on November 2013 SEER data submission, posted to the SEER web site, April Welch HG, Black WC. Are deaths within 1 month of cancer-directed surgery attributed to cancer? J Natl Cancer Inst 2002; 94: United States Census Bureau. Census 2000 Gateway. [online database] Available at: Accessed 27 December In. 10. United States Department of Agriculture Rural-Urban Continuum Codes. [online database] Available at: Accessed 27 December In.

17 Wilt TJ, Shamliyan TA, Taylor BC et al. Association between hospital and surgeon radical prostatectomy volume and patient outcomes: a systematic review. J Urol 2008; 180: ; discussion Nielsen ME, Mallin K, Weaver MA et al. The Association of Hospital Volume With Conditional 90-day Mortality After Cystectomy: An Analysis of the National Cancer Database. BJU Int Tekkis PP, Poloniecki JD, Thompson MR, Stamatakis JD. Operative mortality in colorectal cancer: prospective national study. BMJ 2003; 327: Osler M, Iversen LH, Borglykke A et al. Hospital variation in 30-day mortality after colorectal cancer surgery in denmark: the contribution of hospital volume and patient characteristics. Ann Surg 2011; 253: McCrum ML, Lipsitz SR, Berry WR et al. Beyond volume: does hospital complexity matter?: an analysis of inpatient surgical mortality in the United States. Med Care 2014; 52: Tsai TC, Joynt KE, Orav EJ et al. Variation in surgical-readmission rates and quality of hospital care. N Engl J Med 2013; 369: Rangrass G, Ghaferi AA, Dimick JB. Explaining racial disparities in outcomes after cardiac surgery: the role of hospital quality. JAMA Surg 2014; 149: Osborne NH, Upchurch GR, Jr., Mathur AK, Dimick JB. Explaining racial disparities in mortality after abdominal aortic aneurysm repair. J Vasc Surg 2009; 50: Breslin TM, Morris AM, Gu N et al. Hospital factors and racial disparities in mortality after surgery for breast and colon cancer. J Clin Oncol 2009; 27:

18 Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in hospital mortality associated with inpatient surgery. N Engl J Med 2009; 361: Birkmeyer NJ, Gu N, Baser O et al. Socioeconomic status and surgical mortality in the elderly. Med Care 2008; 46: Pal SK, Hurria A. Impact of age, sex, and comorbidity on cancer therapy and disease progression. J Clin Oncol 2010; 28:

19 19 Figure 1. Forest plots for adjusted odds of 1-month mortality after cancer-directed surgery by significant sociodemographic determinants for each cancer site. (A) Marital Status, (B) Race, (C) Sex, (D) Educational Status, (E) Income, (F) Insurance Status

20 20

21 21

22 22

23 23

24 24

25 Table 1. Baseline clinical and demographic characteristics. Characteristic Suffered 1-month mortality after cancer-directed surgery (N =53,498; 4.8%) Survived beyond 1-month after cancer-directed surgery (N =1,056,738; 95.2%) Age, years, mean (SD) 66 (15) 61 (14) Income, USD, mean 54,200 (13,100) 55,000 (13,200) (SD) Percent that completed 79.5 (7.7) 80.0 (7.6) high school, mean (SD) Sex, N (%) Male 23,385 (43.7) 436,802 (41.3) Female 30,113 (56.3) 619,936 (58.7) Race, N (%) Non-Hispanic 38,907 (72.7) 779,079 (73.7) White Non-white Minority* 14,591 (27.3) 277,659 (26.3) Married, N (%) Yes 28,161 (52.6) 631,844 (59.8) No 25,337 (47.4) 424,894 (40.2) Residence, N (%) Rural 6,714 (12.5) 126,074 (11.9) Urban 46,784 (87.5) 930,664 (88.1) Insurance Status, N (%) ± Insured 22,147 (95.8) 392,863 (96.5) Uninsured 977 (4.2) 14,251 (3.5) Site, N (%) Colorectal 11,129 (20.8) 145, 860 (13.8) Esophageal 314 (0.6) 4,728 (0.5) Thyroid 2,636 (4.9) 60,756 (5.8) Breast 10,790 (20.2) 292,935 (27.7) Pancreatic 891 (1.6) 8,689 (0.8) Endometrial 2,760 (5.1) 58,634 (5.6) Ovarian 1,381 (2.6) 20,284 (1.9)

26 Head/Neck 1,327 (2.5) 26,141 (2.5) Prostate 4,749 (8.9) 149,860 (14.2) Lung 3,475 (6.5) 45,651 (4.3) Liver/IHBD 657 (1.2) 8,610 (0.8) Bladder 4,917 (9.2) 70,071 (6.6) Melanoma 3,038 (5.7) 85,510 (8.1) NHL 2,813 (5.3) 27,856 (2.6) Kidney 2,621 (4.9) 51,153 (4.8) Tumor Stage T1 25,910 (48.5) 558,350 (52.8) T2 12,473 (23.3) 291,025 (27.5) T3 11,137 (20.8) 166,927 (15.8) T4 3,978 (7.4) 40,436 (3.9) All P values comparing characteristics of patients who suffered from 1-month mortality after cancer-directed surgery vs those who did not were. *African American African American, Hispanic/Latino, Asian/Pacific Islander, Native American, or other race County-level data ± Data only available from and numbers shown refer to patients age 65 Abbreviations: IQR = Interquartile Range; IHBD = Intrahepatic Bile Duct; NHL = Non-Hodgkin Lymphoma; N = Number; SD = Standard Deviation; USD = United States Dollar

27 Table 2. Univariable, multivariable, and multivariable sensitivity logistic regression analyses for 1-month mortality after cancer-directed surgery (N = 1,110,236). Characteristic Univariable Analysis Multivariable Analysis A Multivariable Sensitivity Analysis B Multivariable Sensitivity Analysis C Sex OR (95% CI) P Adjusted OR (95% CI) P Adjusted OR (95% CI) P Adjusted OR (95% CI) Female 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) Male 1.10 ( ) Race Non- Hispanic White Non-white Minority* Marital Status 1.10 ( ) 1.11 ( ) 1.07 ( ) 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.05 ( ) 1.13 ( ) 1.13 ( ) 1.12 ( ) Not Married 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) Married 0.75 ( ) Median Household Income Bottom 50 th Percentile Top 50 th Percentile 0.80 ( ) 0.80 ( ) 0.81 ( ) 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) 0.91 ( ) 0.95 ( ) 0.94 ( ) 0.95 ( ) Educational Status Bottom 50 th 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) P

28 Percentile Top 50 th Percentile Age (per year increase) Residence 0.92 ( ) 1.03 ( ) 0.98 ( ) 1.02 ( ) ( ) 1.03 ( ) ( ) 0.21 Rural 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) Urban 0.94 ( ) Insurance Status ± 0.99 ( ) ( ) ( ) Uninsured 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) Insured 0.82 ( ) Tumor Stage ( ) 0.92 ( ) ( ) T1 1.0 (ref) 1.0 (ref) 1.0 (ref) 1.0 (ref) T ( ) T ( ) T ( ) 0.92 ( ) 1.34 ( ) 1.89 ( ) 0.88 ( ) 1.38 ( ) 1.91 ( ) *African American African American, Hispanic/Latino, Asian/Pacific Islander, Native American, or other race 0.89 ( ) 1.08 ( ) 1.30 ( ) A 1-month mortality after cancer-directed surgery defined as death within the same month of diagnosis or in the subsequent month

29 B 1-month mortality after cancer-directed surgery defined as death within the same month of diagnosis (and not in the subsequent month); cancer diagnosis, cancer-directed surgery, and death all occurred in the same month. C Excluding all cases of Non-Hodgkin Lymphoma (N = 30,669) ± Data only available from and analyses limited to patients age 65 Abbreviations: OR = Odds Ratio